Ultraviolet and visible light fused power transmission line discharge detection device

Abstract

The invention relates to the field of power grid safety maintenance and guarantee, and discloses an ultraviolet and visible light fused power transmission line discharge detection device which comprises an unmanned aerial vehicle flying platform, a visible light imaging assembly, a processing unit and an ultraviolet imaging assembly installed on an independent second holder. The processing unit firstly realizes point-cloud-free power transmission tower autonomous identification, tower crown accurate positioning and electronic fence safe descending based on a visible light image; during inspection, target detection and tracking are carried out on power transmission line components in a visible light image, laser ranging information is combined, a space geometric model is solved in real time, and a second holder is actively controlled to enable a visual axis of an ultraviolet camera and a visual axis of a visible light camera to be accurately aligned with the same target; and multi-frame photon accumulation and threshold comparison are carried out on the synchronously collected ultraviolet images. According to the invention, high automation and flexibility of inspection are realized; the problem of fusion precision of a non-coaxial system is solved through active optical axis alignment, and objectivity and reliability of defect judgment are improved through quantitative analysis.

Description

Technical Field

[0001] This invention relates to the field of power grid safety maintenance and protection, specifically to a transmission line discharge detection device that integrates ultraviolet and visible light. Background Technology

[0002] Currently, drone technology has achieved phased results in the application of overhead power transmission line inspection. By equipping itself with a high-definition visible light camera, drones can automatically collect images of power transmission lines and perform intelligent analysis in the cloud. This mode replaces a large amount of traditional manual inspection work, reducing the labor intensity of front-line inspection personnel to a certain extent and improving operational efficiency and safety.

[0003] However, current technology still faces significant bottlenecks in terms of deep autonomy and comprehensive inspection capabilities. At present, autonomous drone inspections generally require pre-construction of 3D models based on laser point clouds, followed by flight path planning. This reliance on high-precision point cloud models not only leads to excessive financial investment in initial data collection and modeling and a lengthy planning cycle, but also makes it difficult to respond quickly to route modifications or terrain changes.

[0004] Secondly, current autonomous inspection operations mostly only target the tower itself, leaving a technological blind spot for detailed inspection of the conductors and ground wires connecting the tower, thus failing to achieve comprehensive coverage of the entire tower and line. Furthermore, the detection methods for live-line inspections are relatively limited, mainly relying on visible light and infrared thermal imaging payloads, which are insufficient for detecting early insulation discharge defects such as corona and arcing. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a power transmission line discharge detection device that integrates ultraviolet and visible light, solving the problem that UAV autonomous inspection can only inspect the tower itself, leaving the conductor inspection area as a blind spot; currently, UAV autonomous inspection still relies on laser point cloud modeling for flight path planning, resulting in excessive upfront investment and a long flight path planning cycle; and UAVs have limited live-line inspection methods during autonomous inspection, relying solely on visible light and infrared payloads for overhead power transmission line inspection, among other problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a transmission line discharge detection device integrating ultraviolet and visible light, comprising a drone flight platform, a visible light imaging component, an ultraviolet imaging component, and a processing unit, characterized in that the visible light imaging component is mounted on the drone flight platform and includes a visible light camera, a first gimbal, and a laser ranging module; the ultraviolet imaging component is mounted on the drone flight platform and includes an ultraviolet camera and a second gimbal; the visible light camera and the ultraviolet camera are non-coaxially arranged; the upper surface of the processing unit is provided with a toothed heat sink; and the processing unit is electrically connected to the visible light imaging component and the ultraviolet imaging component.

[0007] The processing unit is used to perform synchronous rotation calculations, fusing the visible light image acquired by the visible light camera and the ultraviolet image acquired by the ultraviolet camera. The input parameters for the synchronous rotation calculations include:

[0008] The rotation angle of the first gimbal, the rotation angle of the second gimbal, the distance measured by the laser ranging module, and the physical baseline distance between the visible light camera and the ultraviolet camera;

[0009] Through the synchronous rotation calculation, the centers of the visible light image and the ultraviolet image are always aligned with the same target, thereby detecting power transmission line components on the power transmission line.

[0010] Through the above technical solution, this invention solves the technical problem of difficult image alignment caused by visible light cameras and ultraviolet cameras being mounted on different pan-tilt units and having non-coaxial optical paths. The core of this solution lies in the spatial transformation model executed by the processing unit, which can perform pixel-level coordinate mapping in real time based on the dynamically changing geometric relationship between the two cameras. The spatial transformation model can be described by the following equation:

[0011] ;

[0012] in, These are the original pixel coordinates of the ultraviolet discharge signal in the ultraviolet image; These are the pixel coordinates of the final projection of the ultraviolet discharge signal into the visible light image; and These are the pre-calibrated intrinsic parameter matrices for the ultraviolet camera and the visible light camera, respectively. The rotation matrix describes the relative rotation relationship between the two camera coordinate systems. The rotation matrix is ​​calculated from the real-time rotation angle data of the first and second gimbals. The translation vector describing the physical spatial displacement between the optical centers of the two cameras, i.e., the physical baseline distance, is a pre-calibrated fixed value; This refers to the depth value of the target in the ultraviolet camera coordinate system obtained by the laser ranging module; This represents the depth value of the point in the visible light camera coordinate system. This is the inverse of the intrinsic parameter matrix of the ultraviolet camera; The horizontal coordinates in the image; The vertical coordinates in the image are used. The processing unit calculates using a spatial transformation model to precisely superimpose the ultraviolet discharge signal onto the visible light image, achieving accurate visualization and localization of discharge defects.

[0013] Preferably, when fusing the visible light image and the ultraviolet image, the processing unit is further configured to:

[0014] Based on the difference in field of view between the visible light camera and the ultraviolet camera, the ultraviolet image is scaled proportionally.

[0015] The ultraviolet image, after being scaled proportionally, is superimposed onto the visible light image to generate a fused image.

[0016] With the above technical solution, since the field of view of an ultraviolet camera is usually smaller than that of a visible light camera, in order to ensure the spatial scale consistency of the fused image, the processing unit first calculates the ratio coefficient of the field of view of the two cameras, and then uses image scaling algorithms such as bilinear interpolation to adjust the size of the ultraviolet image so that it matches the target area of ​​the visible light image in terms of pixel scale, and then superimposes it.

[0017] Preferably, the processing unit is further configured to determine discharge defects in the ultraviolet image, the process of determining discharge defects including:

[0018] The ultraviolet image is binarized to separate the foreground pixels and background pixels of the discharge region;

[0019] The number of foreground pixels is counted to obtain the photon count value for a single frame.

[0020] The photon count value of a single frame is accumulated over multiple frames to obtain the cumulative photon count value.

[0021] The above technical solution enables quantitative analysis of ultraviolet signals. Binarization converts the original single-channel grayscale ultraviolet image into a binary image containing only pixel values ​​of 0 and 255 by setting a grayscale threshold, effectively separating the discharge signal from background noise. Subsequently, by calculating the total number of foreground pixels with a value of 255, a count value proportional to the number of photons is directly obtained. In one embodiment, to eliminate transient noise interference, a more stable and reliable cumulative photon count value is obtained by summing the photon count values ​​of 32 consecutive frames.

[0022] Preferably, the processing unit is further configured to generate a discharge defect early warning based on the accumulated photon count value, the process of which includes:

[0023] The cumulative photon count is compared with a preset discharge defect determination threshold to determine whether a discharge defect exists, and if a discharge defect is determined to exist, a discharge defect warning is generated.

[0024] Through the above technical solution, the system can automatically determine the severity of the discharge. The discharge defect judgment threshold is set based on a large amount of experimental data and operational experience. When the cumulative photon count exceeds the discharge defect judgment threshold, it is determined that there is a potential equipment insulation defect, and an early warning message is generated.

[0025] Preferably, the processing unit is further configured to:

[0026] The target detection model is run to analyze the visible light image, thereby identifying the type and location of power transmission line components in the image;

[0027] The specific component where the discharge occurred is located by associating the discharge location detected in the ultraviolet image with the location of the power transmission line component identified in the visible light image.

[0028] The above technical solution enables a leap from detecting discharge to locating the specific faulty component. The target detection model can output the bounding box coordinates of the discharging component in the visible light image. The processing unit compares the fused coordinates of the ultraviolet signal with the bounding box. If the ultraviolet signal coordinates are located inside the bounding box of a certain component, the specific component where the discharge occurred is determined to be the source of the discharge fault.

[0029] Preferably, the target detection model is a one-stage target detection model, which includes a multi-level, multi-path feature pyramid to address the loss of contextual information for small targets, and a Siamese auxiliary network to enhance the model's robustness to different lighting conditions.

[0030] The above technical solutions improve the performance of the target detection model in complex power transmission inspection scenarios. The multi-level, multi-path feature pyramid preserves detailed information of small-sized components by fusing feature maps of different depths; the Siamese auxiliary network learns illumination-invariant features by randomly adjusting the brightness and adding noise to the input image during the training phase.

[0031] Preferably, the processing unit is further configured to perform image backlight correction, the process of which includes:

[0032] Run the illumination judgment model to classify the current illumination of the visible light image into preset illumination levels;

[0033] The exposure time and ISO parameters of the visible light camera are adaptively adjusted according to the light level to obtain a properly exposed image.

[0034] The above technical solution solves the problem of unclear imaging of equipment components caused by complex lighting conditions such as backlighting during inspections. The lighting judgment model quantifies the lighting level into 13 levels, ranging from -6 to +6. When the current lighting level is determined to deviate from the normal range, the camera parameters are automatically adjusted to ensure clear imaging of the main components of the power transmission line.

[0035] Preferably, the processing unit is further configured to guide the UAV flight platform to perform precise positioning through a tower crown detection algorithm before performing discharge detection. The tower crown detection algorithm is used to identify the center position of the transmission tower crown and provide feedback to adjust the position of the UAV flight platform so that the UAV flight platform hovers directly above the tower crown.

[0036] The above technical solution enables precise, pointless cloud-based positioning of power poles. The drone only needs to acquire the approximate coordinates of the pole, fly overhead, and then autonomously identify the geometric center of the tower crown using a crown detection algorithm. It then precisely adjusts its own position, providing a precise starting reference point for subsequent autonomous inspection operations.

[0037] Preferably, the processing unit is further configured to:

[0038] After the UAV flight platform completes its positioning based on the tower crown detection algorithm, it is controlled to descend from directly above the tower crown to the side of the tower line.

[0039] During the descent of the drone flight platform, an electronic fence generation algorithm is run to ensure flight safety. The electronic fence generation algorithm is used to calculate the distance between the drone flight platform and the tower in real time, and to perform hovering or alarm when the distance between the drone flight platform and the tower is less than a safety threshold.

[0040] Preferably, the processing unit is further configured to perform anti-duplicate and anti-missed-shot processing on multiple transmission line components identified within the field of view, the process of which includes:

[0041] Assign a unique ID number to each power line component that first appears in the field of view;

[0042] In subsequent image frames, appearance texture feature matching and Kalman filtering are used to continuously track transmission line components carrying ID numbers, so that the shooting task of each component is not repeated or missed.

[0043] This invention provides a power transmission line discharge detection device that integrates ultraviolet and visible light. It has the following beneficial effects:

[0044] 1. This invention performs synchronous rotation calculation through a processing unit. The synchronous rotation calculation integrates the independent gimbal rotation angles of the visible light camera and the ultraviolet camera, the target distance obtained by the laser ranging module, and the physical baseline distance between the two. This solves the technical problem of image alignment difficulties caused by the non-coaxial setting of the visible light and ultraviolet cameras. It achieves the effect that the centers of two different images are always accurately aligned with the same power component target during dynamic inspection, thereby realizing accurate visualization and positioning of ultraviolet discharge faults.

[0045] 2. This invention, through the tower crown detection algorithm and electronic fence generation algorithm built into the processing unit, enables the drone to autonomously fly to the top of the tower crown for precise positioning based only on the approximate coordinates of the tower, and safely descend to the inspection position. This solves the problem that traditional autonomous inspection methods rely heavily on pre-collected high-precision 3D point cloud maps for path planning, and achieves the technical effects of reducing the workload of pre-inspection preparation, shortening the operation cycle and improving deployment flexibility.

[0046] 3. This invention utilizes the image backlight correction function built into the processing unit. This function runs a lighting judgment model to classify the real-time illuminance of visible light images and adaptively adjusts the camera's exposure time and ISO parameters based on the classification results. This solves the problem of poor image quality and unclear target details when UAVs are shooting under complex lighting conditions such as backlight. It achieves the technical effect of prioritizing clear imaging of the main body of power equipment, improving the effectiveness of image data and the accuracy of subsequent intelligent recognition. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0048] Figure 2 This is a three-dimensional schematic diagram of the ultraviolet imaging component of the present invention;

[0049] Figure 3 This is a functional block diagram of the overall system of the present invention;

[0050] Figure 4 This is a flowchart of the point cloud-free autonomous flight process of the present invention;

[0051] Figure 5 This is a flowchart of the ultraviolet and visible light image fusion processing of the present invention;

[0052] Figure 6 This is a flowchart of the ultraviolet discharge defect determination and early warning process of the present invention.

[0053] The components include: 1. Unmanned aerial vehicle (UAV) flight platform; 2. Visible light imaging component; 3. Processing unit; 4. Second gimbal; and 5. Ultraviolet camera. Detailed Implementation

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] To better understand the present invention, the above content will be described in detail below with reference to specific embodiments.

[0056] Please see the appendix Figure 1 - Appendix Figure 6 This invention provides a power transmission line discharge detection device that integrates ultraviolet and visible light. In this embodiment, the processing unit 3 is further configured to execute a method for generating autonomous flight and hovering point mission trajectories without relying on a pre-established three-dimensional point cloud model, so as to guide the UAV flight platform 1 to the predetermined inspection location autonomously, accurately, and safely. This method is executed collaboratively by the following modules:

[0057] The power transmission tower orientation guidance module is used to visually identify and confirm the orientation of a target tower after the UAV flight platform 1 receives the approximate geographical coordinates of the target tower, guiding the UAV to fly towards the target. The core of the power transmission tower orientation guidance module is a power transmission tower detection algorithm, which is a deep learning-based target detection model. To train the target detection model, a dataset containing multiple images was pre-constructed. The images in the dataset were cleaned, filtered, and manually labeled, and divided into two categories: complete towers and partial towers. When the UAV flight platform 1 arrives at the target area, the power transmission tower orientation guidance module uses the onboard forward-looking camera to acquire real-time images. The power transmission tower detection algorithm analyzes the images, identifies the complete tower in the field of view, and calculates the azimuth angle of the tower relative to the UAV's current heading. Based on this azimuth angle, the processing unit 3 generates a heading adjustment command, controlling the UAV flight platform 1 to fly towards the target tower.

[0058] The tower crown precision hovering module is used to precisely correct the position of the UAV platform 1 after it reaches the airspace above the target tower, thus ensuring stable hovering directly above the tower crown. This module solves the positional deviation problem caused by relying solely on Global Navigation Satellite System (GNSS) signals for positioning, which prevents the UAV from accurately positioning itself at the required center position for subsequent operations. The core of the module is a tower crown detection algorithm, also a deep learning model, trained using a dataset of over 5000 visible light images collected directly above the crown of a 500kV transmission tower, with the target category labeled as the tower crown center.

[0059] The specific execution process of the tower crown precision hovering module is as follows: After the UAV flight platform 1 flies above the target tower, the processing unit 3 calls the tower crown detection algorithm to continuously analyze the video stream from the airborne camera. The algorithm detects the pixel coordinates of the center of the tower crown in the image. And calculate the pixel coordinates of the image center point. Deviation vector between Processing unit 3, combining the current flight altitude and camera intrinsic parameters, converts this pixel deviation into a horizontal position correction for the UAV in the world coordinate system. Subsequently, the processing unit 3 generates precise translation control commands to drive the UAV flight platform 1 to make small-range position adjustments. This process is repeated until the magnitude of the deviation vector is less than the preset pixel tolerance threshold. At this point, it is considered that the UAV flight platform 1 has accurately hovered directly above the center of the tower crown.

[0060] The electronic fence safety descent module is used to control the safe descent to the standard inspection height of the tower after the drone flight platform 1 has completed precise positioning above the tower crown. The core of the electronic fence safety descent module is an electronic fence generation algorithm, which dynamically generates a virtual safety boundary for the drone's flight area around the tower to prevent collisions with the tower due to flight control errors or environmental interference.

[0061] The specific execution process of the electronic fence safe descent module is as follows: Before the drone begins its descent, processing unit 3 uses the geographical coordinates of the tower crown center point as a reference to construct a virtual cylindrical or polygonal column on the horizontal plane, based on the maximum radius of the tower structure and with an added safety margin, as a no-fly zone. During the descent, this algorithm performs the following calculations in real time:

[0062] Obtain the real-time 3D coordinates of UAV flight platform 1:

[0063] ;

[0064] This refers to the drone's height above the ground; it also calculates the drone's current horizontal position. Horizontal distance between the tower and the central axis This calculation follows the formula:

[0065] ;

[0066] in:

[0067] This represents the real-time horizontal distance between the drone and the central axis of the tower. The horizontal coordinates of UAV flight platform 1 at the current moment; The horizontal coordinates of the already located center point of the tower crown represent the position of the central axis of the tower.

[0068] The horizontal distance calculated in real time Distance from the preset safety threshold Conduct continuous comparisons.

[0069] when When the condition is triggered, processing unit 3 immediately interrupts the descent command and sends a hovering command to the flight control system of the UAV flight platform 1 or sends an alarm signal to the ground station and the pilot terminal. This mechanism ensures that the UAV maintains a safe distance from the tower throughout the entire vertical maneuver.

[0070] In this embodiment of the invention, processing unit 3 is used to execute a method for fusing ultraviolet and visible light images. The core objective of this method is to accurately register and fuse image information from two physically separate, non-coaxial, and independently moving optical sensors, ultimately generating a composite image that includes a high-resolution visible light background and accurately indicates the location of ultraviolet discharge. This method is executed collaboratively by the following modules:

[0071] The transmission line component target detection module is used to perform real-time analysis of high-definition visible light images acquired by the visible light imaging component 2, automatically identifying and locating various transmission line components present in the images. The core of this module is a deeply optimized stage target detection model, which is based on an improved YOLOv8 architecture. To adapt to the specific characteristics of transmission line inspection scenarios, two key optimizations were performed on the stage target detection model:

[0072] To address the common problem in power transmission lines where small targets in multi-layer convolutional networks are prone to losing contextual feature information due to downsampling, a multi-level, multi-path feature pyramid structure is introduced into the model. This structure establishes additional lateral connections and top-down feature fusion paths between different levels of the backbone network, effectively combining strong semantic information from higher levels with fine spatial details from lower levels. This enhances the model's ability to detect targets at different scales, particularly small targets.

[0073] To overcome the impact of varying lighting conditions on the model's robustness during field inspections, a twin-aided network strategy was employed during training. While the main network was being trained, an auxiliary network with identical structure and shared weights received the same image after data augmentation processing, including random brightness and contrast adjustments and the addition of Gaussian noise. By designing a specific loss function to constrain the consistency of the outputs of the main and auxiliary networks, the model was forced to learn more fundamental target features that are insensitive to changes in lighting, thereby improving stable recognition performance under various lighting conditions. This ultimately enabled the model to identify various types of transmission line components, including insulators, fittings, and conductors.

[0074] The multi-target tracking and management module addresses the challenge of ensuring that each power line component is captured accurately and completely in a continuous video stream when multiple similar or dissimilar components appear simultaneously within the field of view. The core of this module is an anti-duplicate / anti-missed-shot algorithm, the specific execution process of which is as follows:

[0075] Identifier assignment: When a transmission line component is first identified by the target detection module and appears in the field of view, the multi-target tracking and management module assigns a globally unique identifier.

[0076] Feature Extraction and Tracking: In each subsequent frame, the multi-target tracking and management module first extracts the appearance and texture feature descriptors of each target with an assigned identifier using an efficient feature extraction algorithm. Simultaneously, a Kalman filter is used to predict the position, size, and other motion states of each target.

[0077] Data association: Matching newly detected targets in the current frame with tracked targets in the previous frame. The matching process combines the similarity of appearance and texture features with the predictive accuracy of motion states.

[0078] State Update: Successfully matched targets inherit their original identifiers, and the newly detected location is used to update the corresponding Kalman filter state. For unmatched targets, a new identifier is assigned. Tracked identifiers that fail to match any detection results for multiple consecutive frames are marked as having left the field of view. This mechanism maintains stable tracking of the target even when it is briefly occluded or moves rapidly, providing a reliable basis for subsequent shooting task scheduling.

[0079] The cross-platform line-of-view alignment module is used to implement the core function of this invention, namely, to calculate and control the optical axes of the visible light camera and the ultraviolet camera 5, so that they can accurately point to the same point in space. This is the key to solving the field-of-view fusion problem of two non-coaxial, independent gimbal cameras. The core of the cross-platform line-of-view alignment module is a synchronous rotation calculation algorithm, and the real-time input parameters of this algorithm include:

[0080] Real-time rotation angle of the first gimbal housing the visible light camera ;

[0081] Real-time rotation angle of the second gimbal 4 where the UV camera 5 is located ;

[0082] The target distance is measured by the laser ranging module in the visible light imaging component 2;

[0083] The physical baseline vector between the optical centers of the visible light camera and the ultraviolet camera 5 was determined in advance during the device calibration phase.

[0084] Based on the aforementioned input parameters, the cross-platform line-of-sight alignment module dynamically adjusts the angles of one or both gimbals by solving a spatial geometric model, thereby compensating for parallax caused by the baseline vector and target distance. The goal is to ensure that the optical axes of the two cameras intersect at a single point at any given time. When the user or autonomous navigation system controls the visible light camera to point at a target, the cross-platform line-of-sight alignment module calculates in real time the corresponding angle that the ultraviolet camera 5 needs to rotate and drives the second gimbal 4 to complete the compensation movement, and vice versa. Through this closed-loop control, the line-of-sight of the two independent imaging systems is always aligned with the same target.

[0085] The image fusion and overlay module is used to correctly overlay the ultraviolet image information onto the visible light image after the visual axis alignment is completed, generating the final fused image. Because the optical systems of the ultraviolet camera 5 and the visible light camera are different, their field of view differs. To ensure the correct spatial relationship after overlay, the image fusion and overlay module first needs to scale the ultraviolet image proportionally.

[0086] Specific scaling ratio The field of view of the two cameras and Decide:

[0087] ;

[0088] in, It's a tangent function, calculating the tangent value of half the angle of view for each field of view, and then obtaining the result by the ratio of the tangent values ​​of these two field of view angles. That is, the ratio between the two.

[0089] Processing unit 3 according to this ratio The ultraviolet image is scaled using bilinear interpolation or a more advanced interpolation algorithm. Then, the scaled and brightness- or pseudo-color-processed ultraviolet image is superimposed translucently onto the central region of a timestamped visible light image, ultimately generating a fused image that retains high-definition background details while clearly identifying the location and intensity of the ultraviolet discharge signal.

[0090] In this embodiment of the invention, the processing unit 3 is further used to analyze the raw ultraviolet images acquired by the ultraviolet camera 5 to achieve quantitative determination and early warning of discharge defects. This method is executed collaboratively by the following modules:

[0091] The ultraviolet image preprocessing module performs preliminary processing on the raw single-channel grayscale image acquired from the ultraviolet camera 5, thereby highlighting the ultraviolet radiation signal generated by the discharge phenomenon and separating it from background noise. The raw image output by the ultraviolet camera sensor is of U8 data type, with pixel values ​​ranging from 0 to 255 discrete integers. The core processing step of the ultraviolet image preprocessing module is image binarization.

[0092] The ultraviolet image preprocessing module employs either a globally adaptive thresholding method or a fixed empirical threshold. For each frame of ultraviolet image The process involves binarization, which divides each pixel in the image based on its grayscale value and a threshold value. The relationship is reassigned to 0 or 25-5 to generate a binary image. This process can be described by the following formula:

[0093] ;

[0094] in:

[0095] The original ultraviolet image in coordinates The pixel grayscale value at that location; To process the binary image in coordinates Pixel value at; The set binarization threshold; This means if Not greater than the binarization threshold ,Right now Then the ultraviolet image The pixel value at the pixel location in the image is set to 0.

[0096] This operation involves using an image where the grayscale value is higher than a threshold. Pixels marked as 255 constitute the foreground pixels of the discharge area; while grayscale values ​​below or equal to the threshold... The pixels that are not properly counted are suppressed to 0, forming background pixels. This step provides clear, noise-free input data for subsequent photon counting.

[0097] The photon counting and accumulation module is used to perform quantitative analysis on the preprocessed binary image and calculate the photon count value characterizing the discharge intensity. The calculation of the photon counting and accumulation module is based on a fundamental assumption: each independent foreground pixel in the binary image corresponds to the sensor receiving at least one valid photon event at that pixel location.

[0098] The photon counting and accumulation module is executed in two steps:

[0099] Single-frame photon counting: for each frame of a binary image The module iterates through all pixels in the image and counts the total number of foreground pixels with a value of 255. This number is defined as the single-frame photon count value of the image. The calculation formula is:

[0100] ;

[0101] in, Represents the coordinates of all pixels in the image. Perform summation.

[0102] Multi-frame accumulation: Because ultraviolet discharge signals may exhibit instantaneous fluctuations or flickering, relying solely on single-frame photon counts may lead to unstable results or misjudgments. To obtain a smoother and more reliable indicator that better reflects the average intensity of discharge activity, the photon counting and accumulation module employs a multi-frame accumulation strategy. This strategy involves accumulating consecutive frames within a preset time window. Single-frame photon count value Summing these values ​​yields a cumulative photon count. The calculation formula is:

[0103] ;

[0104] in, It is the first in the queue The photon count value per frame of the image. This cumulative value. As the final quantitative basis for measuring the intensity of the discharge, it is passed to the subsequent judgment module.

[0105] The defect detection and early warning generation module is used to make a final determination on the existence of discharge defects based on the cumulative photon count value, and to generate an early warning signal when a defect is confirmed. The working mechanism of the defect detection and early warning generation module is a threshold comparator.

[0106] The defect determination and early warning generation module will use the real-time cumulative photon count value calculated by the photon counting and accumulation module. Compared with the pre-set discharge defect judgment threshold Comparison. Discharge defect determination threshold. It was determined based on factors such as equipment type, voltage level, and environmental conditions, through extensive experimental data and operational experience.

[0107] The logic for the determination is as follows:

[0108] if If the processing unit 3 determines that a significant discharge defect exists at the current detection location, the defect determination and early warning generation module will generate a discharge defect early warning signal. The discharge defect early warning signal can be output in various forms, such as marking the discharge area on the fused image with a highlighted or flashing colored box, or sending a structured alarm message to the ground control station via a data link. The message may contain information such as the coordinates of the defect location, photon count value, and defect level.

[0109] if If the current ultraviolet signal intensity is within the normal range, it is considered not to constitute a discharge defect that requires immediate attention, and no warning is generated.

[0110] Through this quantitative judgment process, this device achieves automated and objective detection and early warning of discharge phenomena in transmission lines.

[0111] In this embodiment of the invention, processing unit 3 is further configured to execute an image backlight correction method. This method automatically adjusts the imaging parameters of the visible light camera during inspection when unfavorable conditions such as insufficient lighting or overexposure exist in the shooting scene, thereby acquiring high-quality images. This method is executed collaboratively by the following modules:

[0112] The illumination level assessment module analyzes the illumination of visible light video images acquired in real time by the visible light imaging component 2, quantifying the current illumination state into preset, discrete illumination levels. The core of the illumination level assessment module is a pre-trained illumination judgment model. The output of the illumination level assessment module is an integer representing the illumination level of the current image. In this embodiment, the range of illumination levels is divided into 13 levels, specifically the set {-6,-5,...,-1,0,1,...,5,6}.

[0113] In a hierarchical system:

[0114] Light level =0 is defined as the normal exposure state and is the target reference benchmark for image quality.

[0115] Light level When the value is negative, it indicates that the image is underexposed to varying degrees. The larger the absolute value, the more severe the underexposure.

[0116] Light level When the value is positive, it indicates that the image is in a state of overexposure to varying degrees; the larger the value, the more severe the overexposure.

[0117] The camera parameter adaptive adjustment module is used to adjust the illumination level based on the illumination level output by the illumination level evaluation module. Through closed-loop feedback, the exposure time and ISO parameters of the visible light camera are adaptively adjusted. The goal of the camera parameter adaptive adjustment module is to drive the illumination level. It converges to 0.

[0118] The camera parameter adaptive adjustment module follows a reverse compensation principle in its adjustment logic. Let the current exposure time of the visible light camera be... The current ISO is The adjusted new parameters are and The adjustment function can be expressed as:

[0119] ;

[0120] The specific behavior of the regulation function is defined as follows:

[0121] When the light level is detected A gain of <0 indicates underexposure of the image. In this case, the camera parameter adaptive adjustment module calculates a gain adjustment amount and increases the camera's total exposure accordingly. Specifically, this involves increasing the exposure time. Or increase light sensitivity Or adjust both simultaneously to make The adjustment range is related to the lighting level. absolute value Relatedly, the more severe the underexposure, the greater the adjustment.

[0122] When the light level is detected A value greater than 0 indicates that the image is overexposed. In this case, the camera parameter adaptive adjustment module calculates a decay adjustment amount and reduces the camera's total exposure accordingly. Specifically, this involves shortening the exposure time. Or reduce the sensitivity Or adjust both simultaneously to make The adjustment range is related to the lighting level. The value is related to the severity of overexposure; the more severe the overexposure, the greater the adjustment required.

[0123] When the light level is detected When =0, it indicates that the current exposure parameters are at their optimal state, and the module will not make any adjustments. .

[0124] Processing unit 3 executes this method through a continuous loop: acquiring an image frame, and having the module evaluate the illumination level. ,like If the light level is not equal to 0, the camera parameter adaptive adjustment module calculates and sends new exposure parameters to the visible light camera, then acquires the next frame and repeats this process. This feedback control loop continues until the image's light level stabilizes near 0, ensuring that the device always obtains a clear, properly exposed image regardless of changes in external lighting conditions.

[0125] In this embodiment of the invention, the overall power supply of the device and the communication link with the UAV flight platform 1 are both implemented through a standardized airborne development interface. The integrated design of the standardized airborne development interface is specifically executed by the following functional modules:

[0126] The power management and distribution module receives a single voltage input from the UAV flight platform 1 and efficiently and stably converts it into a series of different operating voltages required by the various electronic components inside the device.

[0127] The power input terminal of the power management and distribution module is directly connected to the power pin of the airborne development interface of the UAV flight platform 1, and is powered on via an enable pin. When the UAV is powered on and the airborne development interface is enabled, the power management and distribution module starts operating. Internally, it includes a multi-stage, multi-channel power conversion circuit, specifically comprising:

[0128] First-stage DC-DC converter circuit: This circuit efficiently steps down the higher voltage from the UAV to the main voltage rails required by the system, such as 12V or 5V. The DC-DC converter is used to ensure high energy conversion efficiency even under large voltage differences and load currents, thereby reducing heat generation.

[0129] Secondary Low Dropout Linear Regulator Circuit: After the voltage output from the DC-DC converter, a low dropout linear regulator circuit is connected in series in the power supply branch for specific sensitive components. These sensitive components include the computing core board of processing unit 3 and the image sensor of ultraviolet camera 5. The function of the low dropout linear regulator is to perform secondary voltage regulation and filtering on the output of the DC-DC converter, thereby providing a high-purity power supply with extremely low ripple and noise, which is essential for ensuring stable operation of the processor and low-noise imaging by the image sensor.

[0130] The power management and distribution module ultimately outputs multiple stable DC power supplies, which are respectively supplied to the servo motors of the processing unit 3, the ultraviolet camera 5, and the second gimbal 4.

[0131] The multiplex communication and data exchange module establishes and manages all data and control signal links between the device and the UAV flight platform 1, as well as between the various components within the device. To meet different types of interaction needs, the multiplex communication and data exchange module constructs parallel, dedicated communication channels:

[0132] External communication interface: The external communication interface is implemented through a standardized airborne development interface and is divided into two logical channels:

[0133] High-speed data channel: Utilizing the USB 2.0 protocol, the high-speed data channel carries high-bandwidth data stream transmission. Specifically, the real-time visible light video stream acquired by the visible light imaging component 2 and the distance data measured by the laser ranging module are both transmitted at high speed to the processing unit 3 for analysis through this channel.

[0134] Low-speed command and control and telemetry channel: Utilizing the Universal Asynchronous Receiver / Transmitter Protocol (UART), the low-speed command and control and telemetry channel is responsible for exchanging control commands and status information. Real-time flight attitude, GNSS coordinates, altitude, and other telemetry data of the UAV flight platform 1 are sent to the processing unit 3 through this channel; simultaneously, the advanced navigation commands calculated by the processing unit 3 based on algorithms are also transmitted back to the UAV's flight control system via the high-speed data channel.

[0135] Internal communication interface:

[0136] Ultraviolet Image Data Link: The processing unit 3 and the ultraviolet camera 5 are connected via a MIPI camera serial interface. The MIPI camera serial interface is designed for transmitting high-resolution, high-frame-rate raw image data, featuring high bandwidth, low latency, and low power consumption, ensuring that the uncompressed ultraviolet image data captured by the sensor can be delivered to the processing unit 3 completely and in real time for subsequent analysis.

[0137] Gimbal servo control link: Processing unit 3 sends pulse width modulation signals to the servo motor of the second gimbal 4 through the aforementioned general-purpose input / output pins. By precisely controlling the pulse width of the pulse width modulation signal, the rotation angle of the servo motor can be accurately positioned, thereby achieving control of the pitch attitude of the ultraviolet camera 5.

Claims

1. A transmission line discharge detection device integrating ultraviolet and visible light, comprising a UAV flight platform (1), a visible light imaging component (2), an ultraviolet imaging component, and a processing unit (3), characterized in that, The visible light imaging component (2) is installed on the UAV flight platform (1). The visible light imaging component (2) includes a visible light camera, a first gimbal, and a laser ranging module. The ultraviolet imaging component is installed on the UAV flight platform (1). The ultraviolet imaging component includes an ultraviolet camera (5) and a second gimbal (4). The visible light camera and the ultraviolet camera (5) are not coaxially arranged. The upper surface of the processing unit (3) is provided with a toothed heat sink. The processing unit (3) is electrically connected to the visible light imaging component (2) and the ultraviolet imaging component. The processing unit (3) is used to perform synchronous rotation calculation, fusing the visible light image acquired by the visible light camera with the ultraviolet image acquired by the ultraviolet camera (5). The input parameters for the synchronous rotation calculation include: The rotation angle of the first gimbal, the rotation angle of the second gimbal (4), the distance measured by the laser ranging module, and the physical baseline distance between the visible light camera and the ultraviolet camera (5); Through the synchronous rotation calculation, the centers of the visible light image and the ultraviolet image are always aligned with the same target, thereby detecting power transmission line components on the power transmission line.

2. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 1, characterized in that, When fusing the visible light image and the ultraviolet image, the processing unit (3) is also used to: Based on the difference in field of view between the visible light camera and the ultraviolet camera (5), the ultraviolet image is scaled proportionally. The ultraviolet image, after being scaled proportionally, is superimposed onto the visible light image to generate a fused image.

3. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 1, characterized in that, The processing unit (3) is also used to determine the discharge defects in the ultraviolet image, and the process of determining the discharge defects includes: The ultraviolet image is binarized to separate the foreground pixels and background pixels of the discharge region; The number of foreground pixels is counted to obtain the photon count value for a single frame. The photon count value of a single frame is accumulated over multiple frames to obtain the cumulative photon count value.

4. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 3, characterized in that, The processing unit (3) is also used to generate a discharge defect early warning based on the accumulated photon count value, the process of which includes: The cumulative photon count is compared with a preset discharge defect determination threshold to determine whether a discharge defect exists, and if a discharge defect is determined to exist, a discharge defect warning is generated.

5. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 1, characterized in that, The processing unit (3) is also used for: The target detection model is run to analyze the visible light image, thereby identifying the type and location of power transmission line components in the image; The specific component where the discharge occurred is located by associating the discharge location detected in the ultraviolet image with the location of the power transmission line component identified in the visible light image.

6. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 5, characterized in that, The target detection model is a one-stage target detection model. The target detection model includes a multi-level, multi-path feature pyramid to solve the problem of lost contextual information of small targets, and a Siamese auxiliary network to enhance the robustness of the model to different lighting conditions.

7. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 1, characterized in that, The processing unit (3) is also used to perform image backlight correction, the process of which includes: Run the illumination judgment model to classify the current illumination of the visible light image into preset illumination levels; The exposure time and ISO parameters of the visible light camera are adaptively adjusted according to the light level to obtain a properly exposed image.

8. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 1, characterized in that, The processing unit (3) is also used to guide the UAV flight platform (1) to perform precise positioning through the tower crown detection algorithm before performing discharge detection. The tower crown detection algorithm is used to identify the center position of the tower crown of the transmission tower and to adjust the position of the UAV flight platform (1) so that the UAV flight platform (1) hovers directly above the tower crown.

9. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 8, characterized in that, The processing unit (3) is also used for: After the UAV flight platform (1) completes its positioning according to the tower crown detection algorithm, it controls the UAV flight platform (1) to descend from directly above the tower crown to the side of the tower line; During the descent, the UAV flight platform (1) runs an electronic fence generation algorithm to ensure flight safety. The electronic fence generation algorithm is used to calculate the distance between the UAV flight platform (1) and the tower body in real time, and to perform hovering or alarm when the distance between the UAV flight platform (1) and the tower body is less than a safety threshold.

10. The ultraviolet and visible light fusion transmission line discharge detection device according to claim 5, characterized in that, The processing unit (3) is also used to perform anti-duplicate and anti-missed-shot processing on multiple transmission line components identified within the field of view. The process includes: Assign a unique ID number to each power line component that first appears in the field of view; In subsequent image frames, appearance texture feature matching and Kalman filtering are used to continuously track transmission line components carrying ID numbers, so that the shooting task of each component is not repeated or missed.

Images